OpenSEP
Use case

Separate the voices in a podcast from the music and background.

To separate the voices in a podcast from the music and background behind them, upload the episode to OpenSEP and choose Dialogue & background. You get two WAV tracks, the voices and everything else, to preview and download. It is separation only: OpenSEP does not reduce noise or enhance the voices. The Free plan takes files up to 5 minutes; Pro up to 2 hours.

What goes in

  • A podcast episode or another spoken recording with music, a bed or room sound behind the voices.
  • An audio file: MP3, WAV, FLAC, M4A or OGG. An MP4 video works too, and its soundtrack is what gets separated.
  • A stereo or mono mix on every plan. 5.1 and 7.1 files are separated on Pro and Enterprise (downmixed to stereo first); on Free and Standard, export a stereo mix. Transcription takes surround files on any plan.

What comes out

  • Two WAV tracks: the voices, and the music and background behind them. Each keeps your upload's sample rate, and stereo stays stereo.
  • Every voice lands on the same dialogue track. The voices are not split into one track per speaker.
  • Noise that sits under the speech is not removed from the voices: OpenSEP separates, it does not reduce noise or enhance a voice.
  • Traces of music can remain where it overlaps speech, so listen to the voice track before you use it.

Separate an episode

  1. Upload the episode, or the recording you want to work on.
  2. Choose Dialogue & background, not Vocals & instrumental, which is for songs.
  3. Play the voice track alone to check it.
  4. Download both tracks. Results are deleted 3 days after they are produced.
  5. In your editor, put the voice track on its own channel and add new music, or none, under it.

File limits

  • Free: files up to 5 minutes and 128 MiB
  • Standard: files up to 30 minutes and 768 MiB
  • Pro: files up to 2 hours and 3 GiB

Compare every plan

A few useful answers

Before you begin.

More questions? Visit the help center

Does it remove hiss and hum from the voices?

No. Hum, hiss and room echo that sit under the speech stay with the voices. OpenSEP moves the music and background to their own track; it does not repair the voices themselves.

Can I get each host on a separate track?

No. Dialogue & background puts every voice together on the dialogue track. To tell the hosts apart in text, upload the voice track to transcription, which labels each speaker.

Can I process a long episode?

The Free plan takes files up to 5 minutes and 128 MiB, Standard up to 30 minutes and Pro up to 2 hours. A longer episode needs a higher plan, or split it into parts and process each one.

Can I use the separated music again?

Only if you have the right to. Separation does not change who owns the music, so upload recordings you have the right to process.

Try it on your own file.Open your workspace